Show simple item record

dc.contributor.advisorSuhaeni, Cici
dc.contributor.advisorRahman, La Ode Abdul
dc.contributor.authorZuhriyani, Azanti
dc.date.accessioned2026-08-11T06:47:02Z
dc.date.available2026-08-11T06:47:02Z
dc.date.issued2026
dc.identifier.urihttp://repository.ipb.ac.id/handle/123456789/178222
dc.description.abstractRebranding Institut Pertanian Bogor menjadi IPB University memunculkan berbagai opini publik di platform digital yang perlu dianalisis untuk memahami persepsi masyarakat terhadap identitas institusi. Penelitian ini bertujuan menganalisis karakteristik opini publik, mengidentifikasi tema dominan menggunakan latent dirichlet allocation (LDA), serta menganalisis sentimen menggunakan large language model (LLM) berbasis DeepSeek-V4-Flash dengan pendekatan zero-shot prompting. Data penelitian berupa 11.927 komentar dari Google Maps, Instagram, dan YouTube yang diperoleh melalui proses web scraping. Pemodelan topik menghasilkan model terbaik dengan delapan topik pada Google Maps dengan coherence score sebesar 0,44, enam topik pada Instagram dengan coherence score sebesar 0,64, dan tujuh topik pada YouTube dengan coherence score sebesar 0,51. Hasil menunjukkan bahwa Google Maps didominasi pembahasan mengenai lingkungan kampus dan kualitas pendidikan, Instagram mengenai kehidupan mahasiswa dan informasi akademik, sedangkan YouTube menampilkan tema yang lebih beragam sesuai konten video. Evaluasi model LLM pada data sampel memperoleh accuracy sebesar 98,23% dan F1-score sebesar 98,25%. Penerapan model pada seluruh dataset menunjukkan sentimen positif mendominasi sebesar 46,13%, diikuti sentimen netral 39,43% dan sentimen negatif 14,44%. Temuan ini menunjukkan bahwa pendekatan LDA dan LLM efektif dalam menggambarkan tema serta kecenderungan sentimen opini publik terhadap IPB University pascarebranding.
dc.description.abstractThe rebranding of Institut Pertanian Bogor to IPB University has generated diverse public opinions across digital platforms, making it important to understand how the institution is perceived by the public. This study aims to analyze the characteristics of public opinion, identify dominant topics using Latent Dirichlet Allocation (LDA), and examine sentiment using the DeepSeek-V4-Flash Large Language Model (LLM) with a zero-shot prompting approach. The dataset consists of 11.927 comments collected from Google Maps, Instagram, and YouTube through web scraping. Topic modeling identified the optimal models as eight topics for Google Maps with a coherence score of 0,44, six topics for Instagram with a coherence score of 0,64, and seven topics for YouTube with a coherence score of 0,51. The results show that discussions on Google Maps mainly focus on the campus environment and educational quality, Instagram comments emphasize student life and academic information, while YouTube covers a wider range of topics reflecting the content of uploaded videos. The LLM achieved an accuracy of 98,23% and an F1-score of 98,25% on the annotated sample. When applied to the entire dataset, positive sentiment was the most prevalent 46,13%, followed by neutral 39,43% and negative 14,44% sentiments. These findings demonstrate that the combination of LDA and LLM effectively captures both the dominant topics and sentiment patterns in public opinions toward IPB University after its rebranding.
dc.description.sponsorship
dc.language.isoid
dc.publisherIPB Universityid
dc.titlePemodelan Topik LDA dan Analisis Sentimen berbasis LLM pada Opini Publik terhadap IPB University Pascarebrandingid
dc.title.alternativeLDA Topic Modeling and LLM-Based Sentiment Analysis of Public Opinion on IPB University After Rebranding
dc.typeSkripsi
dc.subject.keywordanalisis sentimenid
dc.subject.keywordLDAid
dc.subject.keywordLLMid
dc.subject.keywordopini publikid
dc.subtypeUndergraduate Theses


Files in this item

Thumbnail
Thumbnail
Thumbnail

This item appears in the following Collection(s)

Show simple item record